AIP-C01 study guide: domains, format and a 4-week plan

AIP-C01, the AWS Certified Generative AI Developer – Professional exam, validates that you can integrate foundation models into production applications and business workflows on AWS: RAG, vector stores, agents, prompt management, safety controls, cost and evaluation. It is aimed at developers with two or more years of building production applications and a year of hands-on generative AI work. The exam is 75 multiple-choice and multiple-response questions in 180 minutes, at a Pearson VUE test centre or online.

AIP-C01 at a glance

Details as of September 2026 — confirm on the official exam page before booking.

Who this exam is for

The exam guide describes someone in a generative AI developer role with two or more years of experience building production-grade applications on AWS or with open-source technologies, general AI/ML or data engineering experience, and at least one year of hands-on generative AI implementation. The recommended AWS knowledge covers compute, storage and networking, security and identity, infrastructure as code, monitoring and cost optimisation.

No earlier certification is required. Be honest with yourself about the experience line, though: this is a professional exam, and the scenarios assume you have already shipped and operated a generative AI application. The guide also lists what it does not test: model development and training, advanced ML techniques, and data or feature engineering. It is an integration exam, not a data science exam.

If you are earlier in your path, the [AIF-C01 study guide](/blog/aif-c01-study-guide) covers the foundational exam, and the [MLA-C02 study guide](/blog/mla-c02-study-guide) covers the associate ML engineering exam, which includes model training.

What the exam covers

There are five domains. The first two together make up 57% of the scored content.

Domain 1: Foundation Model Integration, Data Management, and Compliance (31%)

The largest domain, and the one most candidates underestimate because it is broad. It starts with designing a solution and choosing a foundation model against business and technical requirements, including patterns that let you switch models without code changes and designs that stay available during disruption, such as cross-Region inference. The core of it is retrieval: validating and preparing data, designing vector stores and metadata, chunking strategies, choosing embeddings, hybrid search, reranking and query expansion. It closes with prompt engineering as a governed practice, using managed prompt templates, versioning and guardrails rather than strings in code.

Domain 2: Implementation and Integration (26%)

This domain is about building. It covers agentic systems (memory, tool integration, Model Context Protocol servers, stopping conditions and human review steps), deployment choices between on-demand, provisioned throughput and SageMaker AI endpoints, and enterprise integration with API Gateway, Lambda, EventBridge and Step Functions. Expect questions on streaming responses, retries with exponential backoff, fallbacks, model routing and CI/CD for generative AI applications. Know when an asynchronous pattern with a queue is better than a synchronous call.

Domain 3: AI Safety, Security, and Governance (20%)

Here you design layered protection: input and output filtering with Amazon Bedrock Guardrails, grounding responses in a knowledge base to reduce hallucination, and detecting prompt injection and jailbreak attempts. It also covers network isolation with VPC endpoints, least-privilege IAM, PII detection and masking with Amazon Comprehend and Macie, data lineage and audit trails with CloudTrail, and responsible AI practices such as transparency and fairness testing. Questions often ask for the combination of controls that meets a compliance requirement.

Domain 4: Operational Efficiency and Optimization for GenAI Applications (12%)

This domain tests cost and performance thinking specific to generative AI. Learn the levers: prompt and context compression, tiered model selection by query complexity, batch inference, semantic caching, latency-optimised inference and sensible temperature and top-p settings. It also covers observability, tracking token usage, latency and tool calls in CloudWatch and tracing with X-Ray, and troubleshooting failure modes that traditional ML systems do not have.

Domain 5: Testing, Validation, and Troubleshooting (11%)

The smallest domain, but specific. It covers evaluating model outputs for relevance, factual accuracy and consistency, Amazon Bedrock model evaluations, LLM-as-a-judge, RAG and retrieval quality testing, agent evaluation and regression testing of prompts. The troubleshooting tasks centre on context window overflow, API integration errors, prompt regressions and poor retrieval, and on working out which of those is the real cause.

A 4-week study plan

This plan assumes you already build on AWS and have some generative AI experience. Time follows the weights, with extra attention to Domain 1.

Week 1: Domain 1, design and retrieval. Build a small RAG application end to end with an Amazon Bedrock knowledge base, and change one variable at a time: chunking strategy, embedding model, metadata filters, hybrid search. Note how each change affects retrieval. Finish with prompt management and a Domain 1 mixed paper.

Week 2: Domain 1 continued, then Domain 2. Take topic papers on the Domain 1 tasks you scored lowest on. Then turn to agents, tool integration and deployment options. Sketch the integration pattern for three scenarios (synchronous chat, asynchronous document processing, event-driven enrichment) and justify each service.

Week 3: Domains 2 and 3. Finish Domain 2 with resilience and routing. Spend the second half on safety and security: configure guardrails, work out where PII detection belongs in a pipeline, and write down which control answers which risk. Take a mixed paper for each domain.

Week 4: Domains 4 and 5, then full mocks. Cover cost levers, observability and evaluation in the first half. Then sit a full-length mock under exam conditions. Three hours is long, so practise your pacing: roughly two and a half minutes per question with time left to review flagged items. Review, fix weak topics and sit a second mock.

Where candidates lose marks

• Picking an answer that works rather than the one that fits every constraint. Professional scenarios carry several requirements at once: least operational overhead, lowest latency, data residency. Underline each constraint and eliminate options that break any of them.
• Reaching for fine-tuning too early. When the requirement is current or proprietary knowledge, retrieval is the expected answer. Fine-tuning changes behaviour and style, and the guide lists model training as out of scope for the candidate.
• Treating retrieval problems as prompt problems. If the right document never reaches the model, no prompt change fixes it. Troubleshooting questions reward diagnosing chunking, embeddings or metadata before rewriting instructions.
• Misplacing safety controls. Guardrails on output alone do not stop prompt injection or PII reaching the model. Expect questions that need input filtering, output filtering and access control together.
• Multiple-response scoring. AWS gives credit only when every correct option is selected. Read how many answers the question asks for and check each option against the scenario on its own.
• Relying on product names rather than capabilities. Some services the exam guide names, such as Amazon SageMaker Clarify, are no longer open to new customers. Learn what each capability does (grounding, drift detection, agent tracing) so you can recognise it in whichever form the question uses.

Practising for AIP-C01 on CertifyAIx

CertifyAIx has 836 original AIP-C01 questions written against the current exam guide, covering every domain and explained option by option. Our article on [how CertifyAIx questions are written](/blog/how-certifyaix-practice-questions-are-written) explains the method.

• 3 full-length mock exams, 75 questions in 180 minutes, the same as the real exam
• 36 mixed domain papers and 42 topic papers of up to 25 questions each (60 minutes per 25)
• Explanations for every option, with pointers to the relevant AWS documentation
• Progress tracked by domain and topic

Start with the [AIP-C01 practice questions](/certifications/aws-generative-ai-developer-professional-aip-c01).

Frequently asked questions

How difficult is AIP-C01?

It is a professional-level exam, and AWS recommends two or more years of production development plus a year of hands-on generative AI work. The questions are long scenarios with several constraints, so experience of real trade-offs matters more than memorisation.

How long should I prepare for AIP-C01?

With the recommended experience, four to eight weeks of focused study is a realistic range. Allow more if you have not yet built retrieval or agent workloads on AWS, because Domains 1 and 2 assume you have.

Do I need AIF-C01 or another certification first?

No. AWS states that no specific certification is required. AIF-C01 can help with vocabulary, but it does not cover the design and integration depth this exam expects.

Are CertifyAIx questions real AIP-C01 exam questions?

No. Every CertifyAIx question is original and written against the public exam guide. We never use content from a live exam.

What question types are on AIP-C01?

The exam page describes multiple choice (one correct answer out of four) and multiple response (two or more correct answers out of five or more), and the exam guide also lists ordering and matching items, which need every step or pair right to earn credit. Unanswered questions are scored as incorrect and there is no penalty for guessing.

CertifyAIx is independent and not affiliated with Amazon Web Services. AWS Certified Generative AI Developer – Professional is a trademark of its owner. All CertifyAIx practice questions are original.